Trouble ask, training the SVM using SIFT + BOW structures, about 60 eigenvector, trained SVM is used to identify systems, probably how many time can identify as a result, the image is generally 1000 * 1000, thank you all for the great god

CodePudding user response:

It is best to use a large number of actual statistical test images, software and hardware configuration, parameter setting, image characteristics and so on will affect the results